When “Largest Ever” Still Leaves Blind Spots

Cannabis, cognition, and the danger of mistaking variance for damage

When a headline announces that researchers have conducted the “largest ever” study on cannabis and brain function, it arrives with the weight of authority already baked in.

A thousand participants. Brain scans. Cognitive tasks.

It sounds conclusive.

But scientific confidence does not come from scale alone. It comes from what is measured, what is controlled for, and what remains unseen.

The recent JAMA Network Open study by Gowin and colleagues is careful, preregistered, and transparent. It is not sloppy science. It is also incomplete science, in ways that matter deeply to how its findings are interpreted in public discourse.

What the study actually found

Stripped of hype, the paper’s strongest finding is narrow.

Among 1,003 young adults, heavy lifetime cannabis use was associated with lower brain activation during a single working memory task. The effect size was small to moderate. No other cognitive task showed a robust association after correction for multiple comparisons. Diagnosis of cannabis dependence was not associated with brain activation differences.

Recent cannabis use showed transient associations with working memory and motor task activation, but these did not survive statistical correction once education and race were accounted for.

This is not evidence of global cognitive damage. It is an association confined to one domain, under specific conditions.

The authors are clear that the study is cross sectional and cannot establish causality. That honesty matters.

The missing baseline problem

What the study cannot determine is whether cannabis caused the observed differences or whether those differences existed before cannabis use began.

There is no pre-use cognitive baseline. Participants are measured after years of life experience, not before exposure.

The assumed interpretation tends to run in one direction. Cannabis leads to cognitive deficit.

But an equally plausible explanation remains unresolved. Individuals with lower working memory efficiency may be more likely to use cannabis repeatedly.

The dataset cannot distinguish between these possibilities. The authors acknowledge this limitation directly.

The undiagnosed cohort hiding in plain sight

The study classifies participants as “healthy young adults.” That designation excludes major neurological illness. It does not exclude undiagnosed neurodivergence.

There is no screening for ADHD, autism, executive dysfunction, or attentional regulation traits, diagnosed or otherwise.

This omission matters because working memory is the primary outcome where an effect is found. Working memory also shows some of the largest baseline variability across neurodivergent populations.

Adult ADHD and autism are widely underdiagnosed, particularly in women, older adults, and people who masked successfully. In any large adult sample, it is entirely plausible that a substantial minority carry long-standing cognitive profiles that affect working memory performance and neural activation.

Those profiles are invisible in this dataset.

So when lower activation is observed, the study cannot tell us whether cannabis altered the brain or whether cannabis use clustered among people whose brains already worked that way.

Volunteer bias is not neutral

Participation in this study was voluntary and demanding. It required neuroimaging, urine drug screening, and extended cognitive testing.

That filters the population.

People who are distrustful of institutions, concerned about stigma, or opposed to cannabis are less likely to participate. People who are curious about their cognition, motivated to contribute, and comfortable being examined are more likely.

Those traits are not randomly distributed. They overlap strongly with neurodivergent profiles, particularly among undiagnosed adults who have spent years self-monitoring and analysing their internal states.

The study does not examine who declined participation or why. That absence is understandable, but it compounds the interpretive problem.

When variance gets counted twice

Taken together, these factors create a subtle distortion.

Neurodevelopmental traits increase the likelihood of cannabis use.

Those same traits increase the likelihood of volunteering for a study like this.

The study then measures cognitive variability.

That variability is attributed to cannabis exposure.

The same underlying variance is counted once as a predictor and again as an outcome.

No bad faith is required for this to happen. Only unexamined structure.

Funding and framing

To be precise, this study is not funded by pharmaceutical companies. It is supported by NIH and NIDA grants, with funders reporting no role in study design or interpretation.

So the issue here is not covert industry manipulation.

It is ecosystem bias. Research funding systems are structurally oriented toward identifying harms and risks rather than exploring functional substitution or compensatory use. Cannabis is treated as an exposure, not as a response.

The study does not ask what cannabis replaced, what symptoms it alleviated, or whether use was impairing or adaptive for the individual.

That framing choice narrows interpretation without falsifying data.

From paper to headline

Within the paper, language is careful. Associations are clearly labelled. Limitations are acknowledged.

Outside the paper, nuance collapses.

“Cannabis affects the brain” becomes the takeaway.

Technically true. Practically incomplete.

What is lost is the most important question of all.

Whose brains were being measured?

A ToneThread perspective on the error

From a ToneThread and Spectral Binary perspective, the core issue is attribution.

Human cognition exists on a spectrum. Working memory efficiency varies widely across the population. Cannabis use also exists on a spectrum.

When two spectra correlate, binary framing forces one to become the cause and the other the effect.

Spectral framing asks a different question.

Is this signal introduced, amplified, selected for, or merely revealed?

This study cannot answer that question. It never claims to.

The problem begins when others pretend it does.

The responsible conclusion

This paper does not prove that cannabis causes lasting cognitive deficits. It shows a modest association between heavy lifetime cannabis use and working memory task activation in young adults, without accounting for neurodevelopmental variance or baseline cognition.

That is a useful finding. It is not a verdict.

Until research distinguishes between pre-existing variance and induced change, we will continue to mistake diversity for damage, and call it certainty.

Read the article that inspired this post - click here.

To learn more about ToneThread visit www.tonethread.com/studio